Multi-coverage optimal location model for emergency medical service (EMS) facilities under various disaster scenarios: a case study of urban ...
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Nat. Hazards Earth Syst. Sci., 20, 181–195, 2020 https://doi.org/10.5194/nhess-20-181-2020 © Author(s) 2020. This work is distributed under the Creative Commons Attribution 4.0 License. Multi-coverage optimal location model for emergency medical service (EMS) facilities under various disaster scenarios: a case study of urban fluvial floods in the Minhang district of Shanghai, China Yuhan Yang1,3 , Jie Yin1,2,3,4 , Mingwu Ye1,3 , Dunxian She2 , and Jia Yu5 1 Key Laboratory of Geographic Information Science (Ministry of Education), East China Normal University, Shanghai, China 2 Hubei Provincial Key Laboratory of Water System Science for Sponge City Construction, Wuhan University, Wuhan, China 3 School of Geographic Sciences, East China Normal University, Shanghai, China 4 Institute of Eco-Chongming, East China Normal University, Shanghai, China 5 School of Environmental and Geographical Sciences, Shanghai Normal University, Shanghai, China Correspondence: Jie Yin (jyin@geo.ecnu.edu.cn) Received: 3 July 2019 – Discussion started: 18 July 2019 Revised: 15 November 2019 – Accepted: 6 December 2019 – Published: 17 January 2020 Abstract. Emergency medical service (EMS) response is 1 Introduction extremely critical for pre-hospital lifesaving when disaster events occur. However, disasters increase the difficulty of Urban disasters represent a serious and growing challenge. rescue and may significantly increase the total travel time Against the backdrop of urbanization, demographic growth, between dispatch and arrival, thereby increasing the pres- and climate change, the causes of disasters are changing sure on emergency facilities. Hence, facility location deci- and their impacts are increasing. Both natural hazards such sions play a crucial role in improving the efficiency of res- as flash flooding and human-caused accidents such as fires cue and service capacity. In order to avoid the failure of threaten and induce panic in people and cause casualties and EMS facilities during disasters and meet the multiple require- property loss (Kates et al., 2001). In order to deal with emer- ments of demand points, we propose a multi-coverage opti- gencies effectively, a large number of emergency service fa- mal location model for EMS facilities based on the results cilities may be called upon simultaneously. The demands of disaster impact simulation and prediction. To verify this being placed upon emergency services often exceed the re- model, we explicitly simulated the impacts of fluvial flooding sources made available by governments (Liu et al., 2017). events using the 1-D–2-D coupled flood inundation model Furthermore, disasters always require a longer response time FloodMap. The simulation results suggested that even low- than regular incidents due to high traffic flows. A crash on the magnitude fluvial flood events resulted in a decrease in the rescue route may block one or several lanes, resulting in con- EMS response area. The integration of the model results with gestion, significant delays of the emergency vehicles, and po- a geographical-information-system (GIS) analysis indicated tential additional casualties (Dulebenets et al., 2019). There- that the optimization of the EMS locations reduced the delay fore, the maintenance of the efficiency and quality of emer- in emergency responses caused by disasters and significantly gency services during disasters is the key to emergency man- increased the number of rescued people and the coverage of agement. A scientific and pragmatic approach to the choice demand points. of locations and allocations of emergency service facilities reduces traffic congestion and the risk of secondary incidents during an emergency, which, in turn, reduces transport costs and greatly improves the efficiency of rescue services. Published by Copernicus Publications on behalf of the European Geosciences Union.
182 Y. Yang et al.: Multi-coverage optimal location model for EMS facilities under various disaster scenarios Over the last few decades, research into traditional loca- ing: it potentially led to the congestion of emergency ve- tion theory has resulted in a number of models to determine hicles. To solve these problems, Hogan and Revelle (1986) the optimal location of emergency services; the most com- proposed an alternative approach to coverage in their maxi- monly used models are the P-center model (Hakimi, 1964), mal backup coverage models BACOP1 and BACOP2. These the P-median model (Hakimi, 1965), and the covering model models cover each demand point at least once, but the mul- (Brandeau and Chiu, 1989). Among these models, the cover- tiple coverage of different demand points with the same cov- ing model is the most widely investigated and applied model; erage level resulted in a waste of vehicles resources (Ge et the objective of the model is to improve the coverage of fa- al., 2011). Considering that there is usually a limited finan- cilities within a limited time or distance to meet the service cial budget for the provision of emergency services, it is not requirements (Ge et al., 2011). The most common covering feasible to cover all demand points multiple times. models are the location set covering model (LSCM; Tore- The aforementioned traditional location models ignored gas and ReVelle, 1972) and the maximum covering location the impacts of specific disasters, but we suggest that these problem (MCLP) model (Church and ReVelle, 1974). The impacts must be part of any decision regarding the location of focus of the LSCM is to minimize the number of facilities emergency services. Apart from causing casualties, a disas- needed to cover all demand points, but it has been shown ter may also damage emergency facilities; furthermore, dam- to lead to an unequal allocation of facilities or a large in- age to buildings and roads will lead to traffic congestion and crease in costs. Due to these limitations, the MCLP model render emergency rescue more difficult than usual. To avoid was developed to ensure that existing emergency facilities these problems, research has been conducted on choosing were used to obtain the maximum coverage of the demand the location of emergency service facilities in response to points. Drawing upon the LSCM and MCLP model, a num- large-scale emergencies. Jia et al. (2007) defined the main ber of researchers have optimized the associated algorithms characteristics of ideal locations of emergency service facil- in terms of facility workload limits (Pirkul and Schilling, ities as “timeliness”, “fairness”, and “resistance to failure”. 1991), cost (Su et al., 2015), and the level of coverage (Gen- Chen and You (2006) established a multi-objective decision dreau et al., 1997) to solve various practical problems or model for the location of emergency rescue facilities by inte- achieve rescue objectives. Other types of models are suit- grating the MCLP model, the P-median model, and the P- able for location decision problems that do not include time center model. In this integrated model (which focused on or distance restrictions, such as the P-center model and the large-scale disasters), emergency facilities were allocated us- P-median model, where P refers to the number of facilities ing different strategies. Jia et al. (2007) investigated models that need to be built. The P-center model mainly considers for EMS facility location in response to disasters and com- equitable service; it selects P facilities by minimizing the pared three heuristic algorithms (genetic algorithm, location- maximum distance between the demand points and the fa- allocation algorithm, and Lagrange relaxation algorithm) ap- cilities. The P-median model not only takes into account the plicable to emergency scenarios and location models. efficiency of the emergency facilities but also minimizes the After taking the results of previous studies into account, sum of the weighted distance between the demand points and here we describe a novel approach for the optimization of the P facilities (Chen and You, 2006). EMS efficiency under various disaster scenarios. We pro- All of the above models are static in the sense that they do pose a multi-coverage optimal location model, whose out- not consider uncertainties in the emergency service process. put depends on the impact of a disaster and the levels For example, large-scale emergencies are likely to require of demand made on the EMSs. We use a scenario-based high levels of healthcare to the extent that emergency ser- method and geographical-information-system-based (GIS- vice facilities would need to provide transportation to other based) network analysis to quantify the impacts of a disaster facilities that are beyond the immediate area. Furthermore, on the urban EMS response. The coverage level of the de- the limited ambulance resources at any one emergency sta- mand points is determined by the population, the EMS calls tion would restrict the capacity of the emergency medical for help, and other factors that reflect the demand level of service (EMS) when multiple demand points make simulta- the demand points; these factors determine how often the neous requests. Any further demands placed upon the emer- demand point needs to be covered by emergency facilities gency services would cause them to fail, resulting in poten- within a predefined time. The higher the demand coverage tial loss of life. To minimize these fluctuations in an EMS level, the more often a demand point needs to be covered by system, approaches have been proposed that involve multi- the service area of the emergency facilities in a given time pe- coverage models (Moeini at al., 2015). In 1981, Daskin and riod. The main purpose of our location model is to reduce the Stern (1981) put forward their hierarchical objective set cov- probability of delays in the emergency response caused by in- ering model (HOSC), in which they introduced the concept sufficient emergency facilities and resources. The proposed of “multiple coverage of zones”; the objective was to min- model represents a point of reference for the planning and imize the number of necessary facilities such that the de- location of urban emergency facilities. In the following sec- mand was still met and to maximize the coverage of the tions, we provide descriptions of the problems and the design demand points. However, HOSC had one major shortcom- of the optimal location model. We also conduct a case study Nat. Hazards Earth Syst. Sci., 20, 181–195, 2020 www.nat-hazards-earth-syst-sci.net/20/181/2020/
Y. Yang et al.: Multi-coverage optimal location model for EMS facilities under various disaster scenarios 183 – Q1. How do we calculate the coverage level Qi at each demand point i? – Q2. How do we evaluate the risk of disasters for each potential point j and demand point i? – Q3. What are the objectives and constraints for develop- ing an optimized location model based on Q1 and Q2? – Q4. How do we evaluate the applicability of the model? 2.2 Assumptions Figure 1. Qualitative description of model problems. To solve the above problems and simplify the model, we use the following assumptions. of urban fluvial floods in the Minhang district of Shanghai, – A1. All potential points have the same probability of China, to validate this model. accepting EMS calls, and their ability to serve all the demand points throughout the study area is not time- limited. 2 Multi-coverage optimal location model design – A2. During a disaster, each emergency facility has the 2.1 Problem description same service capacity and the same number of ambu- lances. Limited EMS resources face increasing demands as the risk of wide-scale and complex urban disasters increases. Previ- – A3. During a disaster, the closer the EMS is to the source ous models have not considered the probability of failure of of the disaster, the higher the probability is that the fa- EMS facilities, in particular those housing ambulances, nor cility will be unable to respond. have they taken into account possible limited access by EMS to vulnerable demand points. Hence, two problems need to – A4. During a disaster, the closer the EMS is to the source be addressed: (1) the need for quick response times sug- of the disaster, the greater the requirements placed upon gests that EMSs should be located close to potential disaster it from any demand point. points so that a high-risk area can be served simultaneously 2.3 Mathematical model by many EMSs, and (2) the closer to the potential disaster points, the higher the possibility of EMSs being affected by In accordance with the aforementioned description and as- the disaster, and the lower the service capacity, the greater sumptions, a multi-coverage optimal location model is de- the distance should be (Fig. 1). Based on these problems, veloped. In the disaster scenario used for the model, it is as- in this study, we propose and formulate a disaster-scenario- sumed that each emergency facility has the same number of based planning and optimal location model that considers ambulances and quality of service, and we must maximize the multi-coverage of zones. The coverage is dependent on the number of people it can serve within the specified time. the demand level of the demand points (high demand with In order to simplify the model and optimize the algorithm, high coverage requires more ambulances at the same time). we use the 0–1 integer programming method. In our work, we specifically consider flooding; the location The model index sets are as follows: plan should result in improvements in the efficiency of the response and reduce the risk to EMS of flash-flood disasters. – I is set of demand points indexed by i ∈ I = We present the objective of the proposed model and de- {1, . . . , i, . . . , m}. scribe the problems encountered during the development of the model. The objective of the model is to serve the largest – J is the set of potential emergency medical facilities in- number of people in a region with EMSs. Let J be the set dexed by j ∈ J = {1, . . . , i, . . . , n}. of potential emergency facilities, let I be the set of the de- – tij is the time needed for an ambulance to travel from mand points in the study area, and let F (0 < F < J ) be the emergency medical facilities j to demand point i. number of optimal facilities. We consider the risk of a disas- ter at the potential emergency points and the demand points – X is the number of demand points that can be covered separately and arrange the station locations according to the by the service area of the emergency facilities within a coverage level and disaster risk level of each demand point i. specified time. In simple terms, the model solves the following problems. – T is the limit of the emergency response time. www.nat-hazards-earth-syst-sci.net/20/181/2020/ Nat. Hazards Earth Syst. Sci., 20, 181–195, 2020
184 Y. Yang et al.: Multi-coverage optimal location model for EMS facilities under various disaster scenarios – F is number of EMS facilities that need to be built. Equation (5) guarantees that X demand points will be cov- ered within at least T minutes: – Qi is the coverage level of demand point i, meaning that point i should be covered by emergency facilities at zij ≤ yj (∀i ∈ I ; ∀j ∈ J ). (6) least Qi times within a specified time. Equation (6) means that the service point can be serviced – wi is the number of people represented by demand only when this facility is selected: point i. zij ∈ {0, 1}, – mi is the disaster risk level of demand point i. xi ∈ {0, 1}, (7) – pj is the resistance level to the disaster of potential yj ∈ {0, 1}. point j . Equation (7) defines the domains of the decision variables. – xi is binary value, which is equal to 1 if demand point i 2.4 Coverage level analysis is covered; otherwise, it is 0. The model design indicates that the proposed model is based – yj is binary value, which is equal to 1 if an emergency on a goal programming algorithm to optimize the location of medical facility is available; otherwise, it is 0. the EMS facilities based on the existing data and actual sit- uation, the coverage level Qi of each demand point and the – zij is binary value, which equal to 1 if demand point i disaster risk level of the demand points (mi ) and potential fa- is covered by an eligible facility j ; otherwise, it is 0. cilities (pj ). In this section, we propose a new method to es- The overall objective of the model is to rescue the maxi- timate the coverage level that depends on the demand level of mum number of people in a specified time (Question Q3), as the demand point i. Under normal conditions, the demand for shown by the following equation: EMS in one region is mainly related to population attributes such as age distribution and population densities, and areas m X X n of high population densities have a high probability of med- max(z) = mi wi zij pj . (1) ical emergencies. The surrounding conditions also affect de- i=1 j =1 mand, for example, traffic conditions and the presence of reg- ular medical services (such as hospitals). Therefore, in this To keep construction costs under control, the number of study, we analyze the demand level based on these related emergency facilities should be limited. Emergency facilities factors (labeled as evaluation indicators (A)) and evaluate the cannot be built in areas at risk of inundation, and the coverage probability of the demand point calling for help within a pre- rate should be ensured within a specified time. Therefore, the defined time. We then calculate the demand level of every following constraints are added to the multi-objective func- point that is affected by these factors considering the results tion: in terms of the coverage level, i.e., how many times point i n X should be covered by the service area of the emergency fa- yj = F (∀j ∈ J ; 0 < F < J ). (2) cilities. Let A(A = {A1 , A2 , . . . , An ) be the set of indicators j =1 that may affect the coverage level. In order to eliminate the influence of dimension and magnitude and improve the ac- Equation (2) indicates that F emergency facilities should be curacy of the model, the range normalization method is used selected from the potential facilities for emergency services: to convert the original data into the range [0, 1]: n X Ani − min (An ) zij 1/pj ≥ xi Qi pj 6 = 0; ∀i ∈ I ; ∀j ∈ J . (3) A ni = , (8) i=1 max (An ) − min (An ) where Ani represents the normalized index of the indicator Equation (3) ensures that the multiple coverage requirements set A. of the demand points must be met under different disaster These indicators determine the coverage level of demand, scenarios and the resistance level pj to a disaster of potential and they have a certain weight: point j cannot be 0: Qi = INT αA1i + βA2i + · · · + εAni + 1 , (9) tij ≤ T (∀i ∈ I ; ∀j ∈ J ). (4) where αβ· · ·ε represent the weights of thee different indica- Equation (4) ensures that the emergency response time of tors, i.e., their relative contributions to the estimated demand. each ambulance cannot exceed T minutes: The coverage level Qi is then determined by increasing the integers; the results represent the number of times this point xi ≥ X(∀i ∈ I ). (5) needs to be covered by the emergency facilities. Nat. Hazards Earth Syst. Sci., 20, 181–195, 2020 www.nat-hazards-earth-syst-sci.net/20/181/2020/
Y. Yang et al.: Multi-coverage optimal location model for EMS facilities under various disaster scenarios 185 2.5 Disaster risk level analysis city because the water may spread quickly and cover large areas (Green et al., 2017). There are currently 12 emergency Events such as floods, earthquakes, and mudslides can ad- stations in different blocks of this district, and most stations versely affect surrounding buildings and traffic and have seri- are located downtown in the densely populated areas (Fig. 3). ous impacts on EMS. Not only is there is a high probability of Statistical data of the 2017 Shanghai Emergency Center indi- casualties in the disaster source area, which creates high de- cate that the number of EMS calls in 2017 exceeded 40 000 mand for EMS, but the disasters may cause road damage and and the average emergency response time was about 15 min. traffic congestion, making rescue more difficult than usual When large-scale flooding occurs, the emergency response and delaying the emergency response. In order to achieve the efficiency is greatly affected due to the inundation of the road model goal, we analyze the disaster risk level of the demand network. Therefore, we considered a fluvial flood to be a dis- points and potential emergency points and classify the disas- aster scenario for applying the EMS location model. ter level according to the distance of the emergency services from the source of the disaster. For a disaster risk level mi 3.2 Flood impact analysis of demand point i, the closer the point is to the location of the disaster, the higher the risk level and the probability of In order to assess the inundation area and depth following emergency calls for rescue are. For the disaster risk level of fluvial flooding disasters in the study area, we used a 1-D–2- the potential facility j , the closer the facility is to the dis- D coupled flood inundation model named FloodMap (Yu and aster source, the more serious the impact on the facility is, Lane, 2006a, b) to simulate the inundation scenarios of flu- making its location unsuitable for an emergency facility. We vial flooding in various return periods; this model combines express this indicator of the alternative point as the disaster the 1-D solution of the Saint-Venant equations of river flow resistance capacity level pj ; therefore, the disaster resistance with a 2-D flood inundation model based on raster data to of the potential facilities increases with their distance from solve the inertial form of the 2-D shallow water equations. the disaster source. The model is tightly coupled by considering the mass and momentum exchange between the river flow and floodplain inundation, and it is used to simulate the flood process and 3 Case study extract flood potential maps. FloodMap has been applied in For the case study, we choose Minhang district, Shanghai, several different environments and is the mainstream numer- China, as the study area and apply the proposed location ical simulation model used for flood scenarios (Yin et al., model to the optimization of the EMS station distribution 2013b, 2015). We used the FloodMap model to simulate the during the fluvial flooding hazards of Huangpu River based inundation area and depth following fluvial flooding for var- on the data of the Shanghai Emergency Center. The logi- ious return periods (100-year and 1000-year periods) in the gram related of the methodology is shown in Fig. 2. Huangpu River basin in the 2010s, 2030s, and 2050s (Fig. 4). The research data sources include the Shanghai 2013 Trans- 3.1 Study area portation (Gaode) navigation GIS dataset, Shanghai public service facility data, a Shanghai 50 m resolution digital ele- Minhang district is located in Shanghai in China, covers an vation model (DEM), and basic GIS data. area of approximately 372.56 km2 , and is located near the After obtaining the flood scenario simulation results, we Huangpu River. There are nine towns and 514 communi- used various (GIS) tools (e.g., the Spatial Analysis function ties with about 253.4 million residents in the district. The in combination with the Network Analysis function) to as- Huangpu River runs through the entire area, and its river net- sess the impacts of flooding on the EMS response of the ex- work consists of more than 200 rivers, making the study area isting emergency stations. We used the Shanghai Gaode GIS a high-risk area for fluvial flooding. In recent years, due to road network data and the 2017 EMS calls for help data in sea level rise and urban land subsidence, the risk of storm the Minhang district obtained from the Shanghai Emergency surges and floods in the area surrounding to Huangpu River Center. We used five levels for the road speed limit based on has increased (Yin et al., 2013a). Part of the Minhang dis- the People’s Republic of China Technical Standard of High- trict is in the center of Shanghai and has a complex road net- way Engineering (JTG B01-2003) (China Construction Stan- work and dense population; long-term human activities have dard Highway Committee, 2004). Our assessment includes a caused the natural river flow to decrease and the impervious network-based spatial analysis method using the road net- surface areas in the urban areas to increase, making the lo- work data to derive areas that can be reached from an EMS cation highly vulnerable to pluvial floods and fluvial floods. station within a certain timeframe. This method is widely In addition to causing casualties and damaging emergency used in route planning (e.g., via Google Maps navigation) facilities, flood inundation causes damage to buildings and and considers speed limits, road hierarchy, one-way traffic, roads, results in traffic congestion, and complicates emer- and other restrictions in the road networks; this method is gency rescue by delaying the emergency response. Flooding used by network analysis function in the ArcGIS 10.2 soft- causes additional disruption to emergency responders in the ware (New Service Area). Given that the response time is www.nat-hazards-earth-syst-sci.net/20/181/2020/ Nat. Hazards Earth Syst. Sci., 20, 181–195, 2020
186 Y. Yang et al.: Multi-coverage optimal location model for EMS facilities under various disaster scenarios Figure 2. Logi-gram of the multi-coverage optimal location model. Figure 3. Location of the study area. the usual standard by which the efficiency of emergency res- to reduce this to 12 min by 2020. We have therefore used re- cue is assessed, we divided the service area by using the sponse times of 8, 12, and 15 min to divide the EMS service ambulance travel time. In terms of the response time limit area (Yin et al., 2019). In terms of emergency management, for ambulances, 8 min is usually regarded as the standard for when fluvial flood disasters occur, roads near rivers become a medical emergency (Pons and Markovchick, 2002). How- inundated, leading to traffic congestion and road closures, ever, the EMS calls and rescue data from the Minhang district which affect ambulance response times. The failure part of in Shanghai in 2017 indicated that the average medical emer- the transport infrastructure would have the most significant gency response time was about 15 min, although the goal is effects on access to specific locations and the EMS system Nat. Hazards Earth Syst. Sci., 20, 181–195, 2020 www.nat-hazards-earth-syst-sci.net/20/181/2020/
Y. Yang et al.: Multi-coverage optimal location model for EMS facilities under various disaster scenarios 187 Figure 4. Emergency station service areas in the Minhang district under different flood scenario simulations. performance (Albano et al., 2014). Studies have shown that shows the impact on the area serviced by each station for the when road inundation reaches a depth of 30 cm, the roads different flood scenarios. The y axis is the ratio of the ser- become impassable to vehicles (Yin et al., 2016; Green et vice area before and after the disaster: the lower the ratio, the al., 2017). We have, therefore, used an inundation depth of greater the decrease in the service area due to fluvial flood- 30 cm as the road closure restriction for vehicles; we used ing. About half of the stations are affected by the disaster, and the same depth to define the area that cannot be accessed by their service areas have decreased by more than 10 %. The vehicles (the “barrier area”) in our GIS service area analy- starting point for our simulations is the distribution of the ex- sis. We used FloodMap to simulate flood scenarios in 2010, isting Minhang district emergency stations. We find that the 2030, and 2050 for two return periods (100-year and 1000- existing EMS distribution is inadequate for any of the flood year periods). We then used the ArcGIS 10.2 network anal- scenarios used in our model. We, therefore, seek to optimize ysis toolbar to simulate the emergency facility service areas the location of the emergency stations in conjunction with the for the different scenarios (Fig. 4). flood scenarios to ensure that the emergency service facilities Figure 4 shows that during a 100-year flood, one emer- can handle the disasters. gency station (Wujing Station) will lose capacity due to in- undation, whereas a 1000-year flood will affect two stations 3.3 Model parameter calculation (Wujing Station and Jiangchuan Station), both of which are located near the middle and lower drainage basin of the We calculated the two major model parameters (coverage Huangpu River and serve a large population. If these two sta- level and disaster risk level) as proposed in Sect. 2 based tions are incapacitated, it will greatly affect the efficiency of on the flooding scenario results described in Sect. 3.2 and medical emergency rescue in the surrounding areas. Figure 5 used actual data for population, EMS calls for help, etc. We www.nat-hazards-earth-syst-sci.net/20/181/2020/ Nat. Hazards Earth Syst. Sci., 20, 181–195, 2020
188 Y. Yang et al.: Multi-coverage optimal location model for EMS facilities under various disaster scenarios Figure 5. Ratio of the service area of emergency stations before and after the disaster under different flood scenarios. first determined the demand points and number of potential for help at each demand point as the influencing factors A1 emergency stations by dividing the study area into units of and A2 , respectively, of the demand coverage level (using representative blocks or grids. We used data on the location Eq. 9) and used the same weights for the two factors as for of the communities in the Minhang district to determine the a special case (α = β = 0.5 × 10). The resulting Qi is the smallest block unit suitable for modeling demand (each com- coverage level, i.e., the number of times that each demand munity represents a demand unit). We used the ArcGIS 10.2 point i should be covered by the emergency stations in the software Calculate Geometry function to calculate the geo- service area within a specified time. The optimization objec- metric center of each community demand unit as a model tives are to prevent delays in the emergency response caused demand point. To determine the location of potential EMS by busy emergency stations during a disaster, and we con- stations, we covered the entire study area. We divided the strained these objectives using Qi . The results of the demand area into grids of a certain length and assumed that every level calculation are shown in Table 1. grid center point was a potential emergency station. Consid- ering the actual conditions in the research area, we divided 3.3.2 Disaster risk level the area into a grid with a cell size of 2 km × 2 km using the ArcGIS 10.2 fishnet analysis tool (create fishnet). In addition, The results of the disaster-scenario analysis indicate that we added the original 12 emergency stations in the Minhang some existing emergency stations are themselves highly vul- district to these potential stations for comparison. There were nerable to fluvial flooding, which would delay or even pre- 514 demand points and 106 potential stations in the study vent their EMS response. At this stage, we must assess the area (Fig. 6). disaster risk at all points before optimizing the locations of the emergency stations. We considered both the disaster 3.3.1 Coverage level calculation risk level of the demand points and potential stations (Ques- tion Q2); a high risk level not only means that this location is The coverage level Qi of the demand points (Question Q1) unsuitable for the location of EMS but also indicates a high depends on the properties of each point. For example, the need for EMS. larger the population, the more EMS stations are required, We used the disaster risk analysis method proposed in and these should be located nearby. By considering the ex- Sect. 2.5. For the demand point risk level mi , the disaster isting data and the general conditions in the study area, we risk level assessment of the potential stations and the de- regarded the population density and the historical EMS calls mand points are classified by inundation depth. Point i in Nat. Hazards Earth Syst. Sci., 20, 181–195, 2020 www.nat-hazards-earth-syst-sci.net/20/181/2020/
Y. Yang et al.: Multi-coverage optimal location model for EMS facilities under various disaster scenarios 189 Figure 7. Risk level for demand points and potential stations. Figure 6. Demand points and locations of potential stations. the inundation area (depth of more than 30 cm) is regarded Analysis function of ArcGIS to calculate the ambulance driv- as completely inundated at the highest flooding risk level; ing time tij from each potential station j to each demand therefore, we use the area with the inundation depth greater point i during the disaster scenario. The model also included than 30 cm as the center and create three 1 km wide buffer the parameters for the construction of 12 stations (F = 12) zones (mi ∈ {1, 2, 3}). The closer a point is to the inundation to ensure that their service area could cover at least 95 % of center, the higher the risk level of the demand points (Fig. 7). the demand points within 8 min (X ≥ 514 × 0.95, tij ≤ 8). In contrast, the risk level of the potential stations pj can be In simple terms, the objective of this model was to determine regarded as the resistance capacity to a disaster; it increases the locations of emergency stations to rescue the largest num- with the distance to the inundated area. Therefore, we use ber of people in 8 min. We used the demand coverage level the center of the inundation area with a depth of greater than parameters and disaster risk level parameters obtained from 30 cm and divide the disaster resistance level into four 1 km the above-mentioned analysis as inputs for the model and wide buffer zones (pj ∈ {0, 1, 2, 3}). Hence pj = 0 means used Lingo10.0 software to solve the model. The computa- that the potential station j is completely inundated and can- tional results are given in Fig. 8. The central urban area of the not be used as an emergency station. Minhang district is less affected by flooding than other areas; therefore, the location of the EMS stations did not change 3.4 Results significantly. However, in the region near the Huangpu River, the optimized emergency stations are located farther away Here we present the results of the proposed multi-coverage from the inundation area than the current stations, indicating optimal location model for the assignment of the Minhang that the station at the optimized location will be less liable district emergency stations during fluvial flooding and dis- to flooding and more likely to remain operational than the cuss the performance of the optimization of the EMS ser- current stations. vices and coverage level. In order to test our model, we run this model based on the worst-case scenario (1000-year flood 3.4.1 Service capacity comparison in the 2050s). We have assumed that vehicles cannot travel through areas with inundation depths greater than 30 cm. We In terms of emergency management, a service area is an in- utilized the origin–destination (OD) matrix in the Network tuitive measure for determining the service quality of emer- www.nat-hazards-earth-syst-sci.net/20/181/2020/ Nat. Hazards Earth Syst. Sci., 20, 181–195, 2020
190 Y. Yang et al.: Multi-coverage optimal location model for EMS facilities under various disaster scenarios Table 1. Demand point coverage level (sub-sample of the demand point data). Point Area Population EMS Population EMS calls Coverage ID (km2 ) calls density (A1) density (A2) level (Qi ) 1 0.1624119 5225 74 32 171.28 455.6315 4 2 0.06345485 3217 44 50 697.46 693.4064 6 3 0.09560105 3137 59 32 813.45 617.148 4 4 0.2068276 5955 89 28 792.10 430.3101 4 5 0.2035748 6451 150 31 688.60 736.8299 5 6 0.1510978 4728 173 31 290.99 1144.95 6 7 1.463531 11 332 273 7742.92 186.5352 2 8 0.6317168 3317 76 5250.77 120.3071 1 9 3.198358 8736 27 2731.40 8.441831 1 10 0.1303969 3970 61 30 445.52 467.8027 4 11 0.1299455 5082 57 39 108.70 438.6454 4 12 0.3076447 4113 123 13 369.32 399.8118 2 13 0.254323 3115 71 12 248.21 279.1726 2 14 0.08798262 4396 51 49 964.41 579.6599 5 15 0.1688578 4294 37 25 429.68 219.1193 3 16 0.1297367 3815 69 29 405.72 531.8465 4 17 2.101426 2801 113 1332.90 53.773 1 18 3.886865 6481 90 1667.41 23.15491 1 19 0.2178247 4066 58 18 666.38 266.2691 2 20 0.3022524 5911 114 19 556.50 377.1681686 3 Max 10 978 496.3425 25 419 608 76 608.25 1870.493324 8 Min 20 271.96894 86 0 25.7722 0 1 gency service facilities and usually reflects accessibility, i.e., the larger the service area, the larger the number of people who can be served by this station. In general, service areas and population are directly related to the transport infras- tructure conditions around the emergency facilities, includ- ing road speed restrictions and road network density. During flooding, the transport infrastructure near the flooded area will be affected, which will change the travel time of the emergency vehicles, thus reducing the area of emergency ser- vice and accessibility of rescue. Therefore, in this context, we used the service area and population as parameters to evaluate the optimization performance of the model (Ques- tion Q4). Using the ArcGIS 10.2 Service Area Analysis tool, we divided the simulated emergency station service area into three response zones (8, 12, and 15 min journeys) under dif- ferent scenarios; we then used the Spatial Join function to calculate the number of people in the service area. The to- tal service area of the emergency stations for the different response times was calculated, and the comparisons of the service capacity for the current stations and optimal stations are shown in Figs. 9 and 10 using the worst-case flooding scenario (1000-year fluvial flood of the Huangpu River in Figure 8. Computational results of the optimal location model. the 2050s) and the no-flooding scenario. The percent coverage is expressed as a percentage of the total area and the total population; the results suggest that the optimized locations of the emergency stations obtained Nat. Hazards Earth Syst. Sci., 20, 181–195, 2020 www.nat-hazards-earth-syst-sci.net/20/181/2020/
Y. Yang et al.: Multi-coverage optimal location model for EMS facilities under various disaster scenarios 191 Figure 9. Performance comparison of service areas in different scenarios. Figure 10. Service capacity comparison. www.nat-hazards-earth-syst-sci.net/20/181/2020/ Nat. Hazards Earth Syst. Sci., 20, 181–195, 2020
192 Y. Yang et al.: Multi-coverage optimal location model for EMS facilities under various disaster scenarios Figure 11. Comparisons of the average coverage value. Figure 12. Comparisons of the coverage level. by the model provided improvements in the service capacity validate our model results (Question Q4). We combined the over that of the original stations in both the no-flooding and service areas of all emergency stations and used the Spatial extreme flooding scenario based on the 8 min emergency re- Join tool in ArcGIS 10.2 to calculate how many times every sponse time. In the no-flooding scenario, the coverage of the demand point would be covered in 8, 12, and 15 min during service area increased by about 5.5 %, and for the worst-case the no-flooding and the worst-case flooding scenarios. We flooding scenario, the increase was 8.4 %. (Fig. 10); the num- then compared the average values (Fig. 11). ber of people with access to emergency services increased by Figure 11 shows that the average coverage value improved almost 250 000 (10 % increase). These results indicate that after optimization in both scenarios. Specifically, the average the optimization model increased the service capacity for al- coverage value for the no-flooding scenario is slightly higher most all response times and that the performance is best for (about 10 %). The improvement in the average coverage the 8 min response time. value for the no-flooding scenario was greatest for the 12 min response time, i.e., an increase of 6.8 %. For the worst-case 3.4.2 Coverage level performance flooding scenario (1000-year fluvial flood of the Huangpu River in the 2050s), the improvements were more significant: A combination of limited vehicle resources, vulnerable trans- the coverage of the 15 min response time increased by more port infrastructure, and high requirements of the demand than 1 (18.4 %), indicating that, on the average, each demand points during a disaster inevitably place emergency services point can be served by one additional EMS stations within under great pressure. If one demand point is covered by 15 min. These results indicate that using model optimiza- only one emergency station, the limited number of ambu- tion for locating emergency stations greatly improved the lances would soon affect the provision of services for a large response time of emergency services at the demand points, number of demand points, thereby causing delays in rescue. even in an extreme flood disaster scenario, thereby provid- Therefore, a region with high demand should be covered by ing strong disaster resistance. We also compared the percent- multiple emergency service areas that can operate simulta- age of coverage in 8, 12, and 15 min during the no-flooding neously, especially for high-need demand points. The pro- and the worst-case flooding scenarios (Fig. 12). The percent posed model focuses on multiple coverage levels of demand coverage is expressed as a percentage of the demand points points, and we used the real average coverage value for each in different coverage levels. Figure 12 shows that the cover- demand point in a specific time as an important indicator to age level of interval 5–8 is significantly greater for the 8 min Nat. Hazards Earth Syst. Sci., 20, 181–195, 2020 www.nat-hazards-earth-syst-sci.net/20/181/2020/
Y. Yang et al.: Multi-coverage optimal location model for EMS facilities under various disaster scenarios 193 response time and 12 min response time, while that of in- risk factors such as the vulnerability of buildings to evaluate terval 0–3 was significantly decreased; these results indicate the level of disaster risk quantitatively, and we will take into that the model can improve the demand points which have account the terrain and construction cost of the potential lo- a low coverage level for a short response time. In addition, cations. we also found that the optimized coverage level is almost the Lastly, the location of urban emergency service facilities same for the 8 min response during the no-flooding or the has always been an important focus in urban planning. Lo- worst-case flooding scenarios, indicating that extreme fluvial cation selection should consider a variety of factors, and the flooding has little impact on EMS emergency response. ability to respond to disasters should also be considered. In From these results, we can see that stations whose loca- this study, we divided the area into grids with a cell size of tions are determined using the proposed method will have 2 km × 2 km and assumed that every grid center point was a a greater capacity to meet the requirements of the demand potential emergency station; the grid division will affect the points. This reduces the occurrence of “failures” and “in- efficiency of the model and the accuracy of the results. The sufficiency” of emergency stations during disasters, thereby finer the scale, the higher the accuracy but the greater the shortening emergency response times and reducing the loss computational complexity. Therefore, in future research, we of life and property. will consider a multi-scale division that takes into account the population density. In this study, we used a fluvial flooding disaster as an ex- 4 Conclusions ample to analyze the impact of disasters and to evaluate the model. However, the risks faced by cities are not only fluvial This study focused on the optimization of the EMS station lo- floods but also other major events such as earthquakes, mud- cations to ensure efficient emergency medical response in flu- slides, and pluvial floods. In addition, the evacuation plan of vial flood disaster scenarios and the prevention of accidents the population exposed to these hazards should be considered due to emergency response delays and failure of stations. Af- (Alaeddine et al., 2015). Future research should comprehen- ter analyzing the existing location models, we discussed the sively consider a variety of these hazards, conduct risk as- reasons for using multi-coverage plans to improve disaster sessments of the study area quantitatively, and select the lo- emergency resistance instead of traditional location models. cation of urban emergency facilities according to different In addition, since there are various disaster scenarios, we also geographical conditions to improve the efficiency of emer- considered the different damage levels in various areas us- gency response. ing disaster-scenario simulations. The proposed model is an objective programming model with the goal of serving the largest number of people in a specified time during a disas- Data availability. The raw and processed data from the co-authors’ ter. For the case study, we investigated the Minhang district in research findings cannot be shared at this time, as these data are also Shanghai, China, and conducted computational experiments part of the ongoing research. based on real-world data from the Shanghai Emergency Cen- ter. We used the service area and the average coverage level as parameters to evaluate the model performance. The model Author contributions. YY and JY initiated and led this research. YY designed the model, analyzed the performance of this model, results showed that the optimized EMS locations had a wider and wrote the paper. JY provided EMS history records and flood- service range for 8 min response time and a larger number of ing inundation simulation. MY dealt with the road data. DS and people were served; the coverage level was also improved. JY helped in analyzing emergency data. The coverage level of some of the existing stations changed greatly after the disaster, whereas the optimized location re- sults showed that the service level before and after the disas- Competing interests. The authors declare that they have no conflict ter was almost the same. Both parameters indicated that the of interest. proposed multi-coverage location optimization model is well suited to model the emergency response to flood disasters and to conduct site selection of urban emergency facilities. Acknowledgements. This work was supported by the National Some aspects of the model could be improved to obtain Key Research and Development Program of China (grant a more robust solution. First, in the case study, we did not no. 2018YFC1508803), the Major Project of the National Social conduct a quantitative assessment of the effect of the disaster Science Fund of China (grant no. 18ZDA105), the National Natu- risk level on the emergency response, but we evaluated the ral Science Foundation of China (grant nos. 41871164, 41601568, disaster risk level by using the buffer distance to the source and 51761135024), the Humanities and Social Sciences Project of the Ministry of Education of China (grant no. 17YJAZH111), of the disaster, which is a subjective approach. Second, since the Key Project of Soft Science Research of Shanghai (grant this was a theoretical analysis, our model did not consider no. 19692108100), the Fundamental Research Funds for the Cen- whether the terrain or other basic conditions were suitable for tral Universities (grant nos. 2018ECNU-QKT001 and 2017ECNU- the EMS facilities. In future studies, we will consider disaster www.nat-hazards-earth-syst-sci.net/20/181/2020/ Nat. Hazards Earth Syst. Sci., 20, 181–195, 2020
194 Y. Yang et al.: Multi-coverage optimal location model for EMS facilities under various disaster scenarios KXK013), and the Hubei Provincial Key Laboratory of Water Sys- Green, D., Yu, D., Pattison, I., Wilby, R., Bosher, L., Patel, R., tem Science for Sponge City Construction (grant no. 2019-02). Thompson, P., Trowell, K., Draycon, J., Halse, M., Yang, L., and Ryley, T.: City-scale accessibility of emergency responders op- erating during flood events, Nat. Hazards Earth Syst. Sci., 17, Financial support. This research has been supported by the Na- 1–16, https://doi.org/10.5194/nhess-17-1-2017, 2017. tional Key Research and Development Program of China (grant Hakimi, S. L. J. O. R.: Optimum Locations of Switch- no. 2017YFE0100700), the National Natural Science Founda- ing Centers and the Absolute Centers and Medians tion of China (grant no. 41871164), the Humanities and Social of a Graph, Oper. Res. Manage. Sci., 12, 450–459, Sciences Project of the Ministry of Education of China (grant https://doi.org/10.1287/opre.12.3.450, 1964. no. 17YJAZH111), and the Hubei Provincial Key Laboratory Hakimi, S. L. J. O. R.: Optimum distribution of switch- of Water System Science for Sponge City Construction (grant ing centers in a communication network and some re- no. 2019-02). lated graph theoretic problems, Oper. Res., 13, 462–475, https://doi.org/10.1287/opre.13.3.462, 1965. Hogan, K. and Revelle, C.: Concepts and Applications of Backup Review statement. This paper was edited by Sven Fuchs and re- Coverage, Manage. Sci., 32, 1434–1444, 1986. viewed by Kamal Serrhini and one anonymous referee. Jia, H. Z., Ordonez, F., and Dessouky, M.: A model- ing framework for facility location of medical ser- vices for large-scale emergencies, IIE Trans., 39, 41–55, https://doi.org/10.1080/07408170500539113, 2007. Kates, R. W., Clark, W. C., Corell, R., Hall, J. M., Jaeger, C. C., References Lowe, I., McCarthy, J. J., Schellnhuber, H. J., Bolin, B., Dick- son, N. M., Faucheux, S., Gallopin, G. C., Grubler, A., Huntley, Alaeddine, H., Serrhini, K., Maizia, M., and Néron, E.: A spatiotem- B., Jager, J., Jodha, N. S., Kasperson, R. E., Mabogunje, A., Mat- poral optimization model for the evacuation of the population son, P., Mooney, H., Moore, B., O’Riordan, T., and Svedin, U.: exposed to flood hazard, Nat. Hazards Earth Syst. Sci., 15, 687– Environment and development – Sustainability science, Science, 701, https://doi.org/10.5194/nhess-15-687-2015, 2015. 292, 641–642, 2001. Albano, R., Sole, A., Adamowski, J., and Mancusi, L.: A GIS-based Liu, M., Yang, D. P., and Hao, F. X.: Optimization for the model to estimate flood consequences and the degree of accessi- Locations 70 of Ambulances under Two-Stage Life Res- bility and operability of strategic emergency response structures cue in the Emergency Medical Service: A Case Study in in urban areas, Nat. Hazards Earth Syst. Sci., 14, 2847–2865, Shanghai, China, Math. Probl. Eng., 2017, 1830480, 2017, https://doi.org/10.5194/nhess-14-2847-2014, 2014. https://doi.org/10.1155/2017/1830480, 2017. Brandeau, M. L. and Chiu, S. S.: An Overview of Representa- Moeini, M., Jemai, Z., and Sahin, E.: Location and relocation tive Problems in Location Research, Manage. Sci., 35, 645–674, problems in the context of the emergency medical service sys- 1989. tems: a case study, Central Eur. J. Oper. Res., 23, 641–658, Chen, Z. Z. and You, J, X.: A Multi-objective Decision Model of https://doi.org/10.1007/s10100-014-0374-3, 2015. Emergency Rescue Facility Location for Large-scale Emergency Pirkul, H. and Schilling, D. A.: The Maximal Covering Location Incidents, Manage. Sci. China, 19, 10–14, 2006. Problem with Capacities on Total Workload, Manage. Sci., 37, China Construction Standard Highway Committee: JTG B01-2003, 233–248, https://doi.org/10.1287/mnsc.37.2.233, 1991. People’s Republic of China industry standard: highway engi- Pons, P. T. and Markovchick, V. J.: Eight minutes or less: Does the neering technical standards, People’s Communications Publish- ambulance response time guideline impact trauma patient out- ing House, Beijing, 2004. come?, J. Emerg. Med., 23, 43–48, 2002. Church, R. and ReVelle, C.: The maximal covering lo- Su, Q., Luo, Q. Y., and Huang, S. H.: Cost-effective analyses for cation problem, Pap. Reg. Sci. Assoc., 32, 101–118, emergency medical services deployment: A case study in Shang- https://doi.org/10.1007/bf01942293, 1974. hai, Int. J. Prod. Econ., 163, 112–123, 2015. Daskin, M. S. and Stern, E. H. J. T. S.: A Hierarchi- Toregas, C. and ReVelle, C.: Optimal location under time or cal Objective Set Covering Model for Emergency Medical distance constraints, Pap. Reg. Sci. Assoc., 28, 131–143, Service Vehicle Deployment, Transport Sci., 15, 137–152, https://doi.org/10.1007/BF01961457, 1972. https://doi.org/10.1287/trsc.15.2.137, 1981. Yin, J., Yu, D. P., Yin, Z. N., Wang, J., and Xu, S. Y.: Modelling the Dulebenets, M. A., Abioye, O. F., Ozguven, E. E., Moses, R., Boot, combined impacts of sea-level rise and land subsidence on storm W. R., and Sando, T.: Development of statistical models for im- tides induced flooding of the Huangpu River in Shanghai, China, proving efficiency of emergency evacuation in areas with vul- Climatic Change, 119, 919–932, 2013a. nerable population, Reliabil. Eng. Syst. Safe., 182, 233–249, Yin, J., Yu, D. P., Yin, Z. N., Wang, J., and Xu, S. Y.: Multiple sce- https://doi.org/10.1016/j.ress.2018.09.021, 2019. nario analyses of Huangpu River flooding using a 1D/2D coupled Ge, C, J., Wang, X., and Guan, X. J.: A Multi-Covering Model flood inundation model, Nat. Hazards, 66, 577–589, 2013b. and Its Algorithm for Facility Location Response for Large- Yin, J., Yu, D. P., Yin, Z. N., Wang, J., and Xu, S. Y.: Modelling the Scale Emergencies, Oper. Res. Manage. Sci., 20, 50–56, anthropogenic impacts on fluvial flood risks in a coastal mega- https://doi.org/10.1007/s11424-010-8015-y, 2011. city: A scenario-based case study in Shanghai, China, Landsc. Gendreau, M., Laporte, G., and Semet, F.: Solving an ambu- Urban Plan., 136, 144–155, 2015. lance location model by tabu search, Locat. Sci., 5, 75–88, https://doi.org/10.1016/S0966-8349(97)00015-6, 1997. Nat. Hazards Earth Syst. Sci., 20, 181–195, 2020 www.nat-hazards-earth-syst-sci.net/20/181/2020/
Y. Yang et al.: Multi-coverage optimal location model for EMS facilities under various disaster scenarios 195 Yin, J., Yu, D. P., Yin, Z., Liu, M., and He, Q.: Evaluating the impact Yu, D. and Lane, S. N.: Urban fluvial flood modelling using a two- and risk of pluvial flash flood on intra-urban road network: A dimensional diffusion-wave treatment, part 1: mesh resolution case study in the city center of Shanghai, China, J. Hydrol., 537, effects, Hydrol. Process., 20, 1541–1565, 2006a. 138–145, 2016. Yu, D. and Lane, S. N.: Urban fluvial flood modelling using a Yin, J., Jing, Y. M., Yu, D. P., Ye, M. W., Yang, Y. H., and two-dimensional diffusion-wave treatment, part 2: development Liao, B. G.: A Vulnerability Assessment of Urban Emer- of a sub-grid-scale treatment, Hydrol. Process., 20, 1567–1583, gency in Schools of Shanghai, Sustainability-Basel, 11, 1–12, 2006b. https://doi.org/10.3390/su11020349, 2019. www.nat-hazards-earth-syst-sci.net/20/181/2020/ Nat. Hazards Earth Syst. Sci., 20, 181–195, 2020
You can also read